superbigsaw/MansionWorld

MansionWorld

<p align="center"> <img src="teaser.png" width="100%" /> </p>

MansionWorld is a large-scale multi-floor indoor scene dataset for embodied AI research, built on top of the AI2-THOR simulator.

Key Features

• —1,000 interactive multi-floor buildings with 2–10 floors each
• —10,000+ rooms in total across all buildings
• —Diverse building types: Residential (50%), Office (30%), Public (20%)

<p align="center"> <img src="mansionworld_dataset.png" width="80%" /> </p>

Download

The dataset is provided as zip archives for easier downloading:

Dataset Structure

After extracting mansionworld.zip:

mansionworld/
├── {building_type}_{floors}f_{area}_fp{id}#{variant}/
│   ├── floor_1.json      # Floor configuration (doors, walls, objects, rooms)
│   ├── floor_1.png       # Floor plan visualization
│   ├── floor_2.json
│   ├── floor_2.png
│   └── ...
└── ...

Building Naming Convention

Example: office_corporate_hq_6f_300_fp001#1

• —office_corporate_hq: Building type (office - corporate headquarters)
• —6f: Number of floors (6 floors)
• —300: Approximate floor area (m²)
• —fp001: Floorplan template ID
• —#1: Variant index

Usage

For loading MansionWorld scenes and using the Mansion API, please refer to the documentation.

Quick Download

git clone https://huggingface.co/datasets/superbigsaw/MansionWorld

Links

• —📄 Paper: arXiv:2603.11554
• —🐙 Code: AgibotGeneral/MANSION
• —🌐 Project Page: agibotgeneral.github.io/mansion-site

Citation

If you use MansionWorld in your research, please cite:

@misc{che2026mansionmultifloorlanguageto3dscene,
      title={MANSION: Multi-floor lANguage-to-3D Scene generatIOn for loNg-horizon tasks}, 
      author={Lirong Che and Shuo Wen and Shan Huang and Chuang Wang and Yuzhe Yang and Gregory Dudek and Xueqian Wang and Jian Su},
      year={2026},
      eprint={2603.11554},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2603.11554}, 
}

License

This dataset is released under the CC BY 4.0 License.